comfyui-mcp 0.50.19 → 0.50.21

package/dist/orchestrator/panel-tools.js

@@ -6006,6 +6006,20 @@ export function buildPanelToolDefs() {
                 isRetryableRunToNodeStampRace(res, rejection)) {
                 scopeRebuilt = true;
                 try {
+                    // #1050 — SETTLE before re-issuing. The panel refuses this race when the
+                    // graph changed between dispatch and apply, and re-dispatching in the
+                    // same tick lands in the SAME window: a reporter's retry raced
+                    // identically and instantly, so the one re-issue was spent for nothing
+                    // and the run was refused twice with nothing queued.
+                    //
+                    // This is a SETTLE PAUSE, not the mutation-quiescence barrier the report
+                    // asked for. The panel exposes no such signal, and inventing a
+                    // graph-is-quiet reading from out here would be guessing at the
+                    // frontend's state. It buys pending edits a moment to land — exactly
+                    // what the reconnect retry already does for a dropped socket — and
+                    // nothing more. A graph still moving after it (a user actively editing)
+                    // races again and is SURFACED, never re-raced.
+                    await sleep(retrySettleMs());
                     res = await ctx.call(runCmd, 20000);
                     rejection = detectRunRejection(res);
                 }
@@ -6033,9 +6047,11 @@ export function buildPanelToolDefs() {
                     return rejection;
                 return appendToolResultText(rejection, `\n\n(Dispatch history: the first graph_run was refused by the panel's run-to-node ` +
                     `graph-stamp race, which CERTIFIED that nothing was queued, so the scoped run was ` +
-                    `re-issued exactly once. The failure above is that SECOND dispatch; it was not ` +
-                    `retried again. Judge whether anything was queued from that message alone — the ` +
-                    `first dispatch definitely queued nothing.)`);
+                    `re-issued exactly once — after a short pause to let pending graph edits land ` +
+                    `(#1050). The failure above is that SECOND dispatch; it was not retried again. ` +
+                    `Judge whether anything was queued from that message alone — the first dispatch ` +
+                    `definitely queued nothing. Racing AGAIN after the pause means the graph is still ` +
+                    `changing under the run: let the canvas settle, then re-run.)`);
             }
             // Attribute this genuine queue to ourselves so a later panel_run in the
             // same batch recognizes the in-flight job as our own and doesn't false-warn